ABSTRACT
FABIA (factor analysis for bicluster acquisition; Hochreiter et al., 2010) evolved into one of the most successful biclustering methods. It has been applied to genomics, where it identified in gene expression data task-relevant biological modules (Xiong et al., 2014). For example in drug design, first chemical compounds are added to cell lines and then the change in gene expression is measured. The goal is to identify biological modules that are altered by chemical compounds to assess their biological effects (see Chapter 12-Chapter 15 for more details). In another drug design application, FABIA was used to extract biclusters from a data matrix that contains bioactivity measurements across compounds. An entry in the data matrix is the activity level of a compound for a specific bioassay. Biclusters correspond to compounds that are active on similar bioassays, that is compounds that interact with similar proteins, e.g., by binding to them. Activity includes both desired and undesired effects, that is efficacy and side effects (see Chapter 12 for more details). FABIA has been applied to genetics, where it has been used to identify DNA regions that are identical by descent (IBD) in different individuals. These individuals inherited an IBD region from a common ancestor (Hochreiter, 2013; Povysil and Hochreiter, 2014). IBD analysis is utilised in population genetics
to find the common history of populations and when they split. Further IBD mapping is used in association studies to detect DNA regions which are associated with diseases. An IBD region is characterized by nucleotide mutations that were already present in the common ancestor and are now shared between individuals. Therefore, these individuals are similar to one another on these mutations but not on others. This is a perfect setting for biclustering where a bicluster corresponds to an IBD segment. See Chapter 16 for an application of FABIA in genetics.
